<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-1025">
  <Title>A Method for Open-Vocabulary Speech-Driven Text Retrieval</Title>
  <Section position="9" start_page="0" end_page="0" type="relat">
    <SectionTitle>
8 Related Work
</SectionTitle>
    <Paragraph position="0"> The method proposed by Kupiec et al. (1994) and our method are similar in the sense that both methods use target collections as language models for speech recognition to realize open-vocabulary speech-driven retrieval.</Paragraph>
    <Paragraph position="1"> Kupiec et al's method, which is based on word recognition and accepts only short queries, derives multiple transcription candidates (i.e., possible word combinations), and searches a target collection for the most plausible word combination. However, in the case of longer queries, the number of candidates increases, and thus the searching cost is prohibitive. This is a reason why operational speech recognition systems have to limit the vocabulary size.</Paragraph>
    <Paragraph position="2"> In contrast, our method, which is based on a recent continuous speech recognition framework, can accept longer sentences. Additionally, our method uses a two-stage retrieval principle to limit a search space in a target collection, and disambiguates only detected OOV words. Thus, the computation cost can be minimized.</Paragraph>
  </Section>
class="xml-element"></Paper>